Presently, commitUnitOfWork() implementations do both storage engine-specific and non-storage engine-specific commit procedures.

In order to cleanly separate out the generic commit procedures, we could consider separating commitUnitOfWork() into two functions, both called in succession wherever commitUnitOfWork() is presently called. One function would do commit work specific to the storage engine, and would be required for each storage engine to implement, and the other would do generic commit work, and would only be implemented at the RecoveryUnit level.